Week 4 Interactions

Estimating, graphing and interpreting interaction effects

Description

This week we will return to some of the models we fit over the past two weeks and turn our attention to the right-hand side of the regression equations. We will explore interaction effects, which allow for the association between a predictor and an outcome to depend upon the value of another predictor. Understanding interaction effects - and the concept of conditioning more broadly - can help avoid serious misrepresentations and misunderstandings of our data. Some famous statistical “paradoxes” can highlight these dangers, and the lecture will build on these conceptual examples before moving on to questions of model-building. It is technically very simple to fit models with complex interactions, however, their interpretation can be very difficult. As with the logistic regression model we covered in Week 3, we will explore ways to present and visualise results from interaction models, which make their interpretation easier. Exploring interactions also allows us to begin thinking about causality and causal modelling, a topic that we will expand upon in Week 5.

Readings

Statistics

• ROS: Chapters 10-12

Coding

• TSD: Chapter 15

Application

• Österman, Marcus. 2021. ‘Can We Trust Education for Fostering Trust? Quasi-Experimental Evidence on the Effect of Education and Tracking on Social Trust’. Social Indicators Research 154(1):211–33 - (online)

• Ladd, Jonathan McDonald, and Gabriel S. Lenz. 2009. ‘Exploiting a Rare Communication Shift to Document the Persuasive Power of the News Media’. American Journal of Political Science 53(2):394–410. doi: 10.1111/j.1540-5907.2009.00377.x.(published version should be accessible with university login; additional Appendix available here)